Task statement 5.3 · Context management and reliability

Implement error propagation strategies across multi-agent systems

The research system is writing a report on grid-scale battery storage in six European markets. The coordinator sends four topics to the web search subagent: cost trends, manufacturers, deployments and national incentives. The incentives topic needs one query per market, and the queries for Italy and Poland time out. What the report says about those two markets depends on the subagent's reply.

Scenario switchThis section runs on the exam guide's Multi-Agent Research System, the domain's declared second scenario. A coordinator delegates to web search, document analysis, synthesis and report generation subagents, and the guide writes error propagation about exactly that arrangement.
Why it's tested

The tempting fixes all look tidy. Returning empty lists for the timeouts keeps the run green, but the coordinator reads "searched, found nothing", so the report states that Italy has no incentives. Retrying and then returning "search unavailable" is honest, but it drops the failed queries and the four markets that worked, so the coordinator can only repeat everything blind or drop the topic. Stopping the run fails the other way, because one unreachable source then discards three finished topics.

KnowledgeWhat crosses the agent boundary

A subagent runs in its own context window, and only its final message returns to the coordinator. The coordinator never sees the timeouts, the retries or the pages read, so that message is its only evidence about what failed. Section 2.2 marked a single tool's failed result with isError: true and added a category and a retry flag, so that its caller can recover. A subagent's message must carry more, because the coordinator chooses between whole routes: the failure type, what was attempted, the partial results already obtained, and the alternatives the subagent knows of. A generic status such as "search unavailable" says only that something failed, so the coordinator cannot tell a two-market gap from a dead search engine.

  1. Four markets answer; the Italy and Poland queries time out.
  2. The subagent retries both itself, twice each, with backoff.
  3. Both still fail, so it propagates what it could not resolve.
  4. Its final message carries the failure type, the queries attempted, the four markets' results, and an alternative source.
  5. The coordinator reroutes the two markets, or proceeds and has the gap disclosed.

Interactive · four ways to report the same two timeouts

Top to bottom: the coordinator delegates four topics; the incentives topic covers six markets, and Italy and Poland time out; the web search subagent retries locally and still fails; what its final message tells the coordinator; the coordinator's next move; what reaches the final report; and two meters, for markets with incentive data in the report and for missing markets the report discloses COORDINATOR DELEGATES FOUR TOPICS TO WEB SEARCH cost trends done manufacturers done incentives · 6 markets 4 answer · IT, PL time out deployments done local recovery: 2 retries each, backoff WEB SEARCH SUBAGENT · ITS OWN CONTEXT WINDOW IT and PL still time out after the retries. It cannot resolve that alone, so it reports upward. final message to the coordinator WHAT THE COORDINATOR RECEIVES { "status": "success", "results": { DE, FR, ES, PT, "IT": [], "PL": [] } } { "status": "error", "message": "search unavailable" } TimeoutError raised to the top-level handler failure_type: "timeout" · attempted: "incentives IT", "incentives PL", 2 retries each — no sign that two queries timed out — — no failure type, no queries, the four answered markets discarded — — the workflow stops — partial_results: DE, FR, ES, PT · alternatives: national energy agency registries coordinator's next move COORDINATOR Reads six finished searches. There is nothing to recover from. Knows the incentives topic failed, not where or why. It repeats all six queries blind, or drops the topic. Never gets a turn. The run is terminated. Re-delegates IT and PL to the document analysis subagent, pointed at the registries. Italy's registry answers. what reaches the report FINAL REPORT “Italy and Poland offer no national incentives.” Two absences stated as findings. The incentives topic is missing, and nothing tells the reader so. No report. Three finished topics and four answered markets are discarded with it. Incentives for 5 of 6 markets. Coverage note: Poland's source unreachable, so its figures are absent, not zero. INCENTIVE DATA IN THE REPORT · MARKETS 4 of 6 0 of 6 5 of 6 MISSING MARKETS THE REPORT DISCLOSES 0 of 2 0 of 6 1 of 1 Illustrative run: six incentives queries, four answer, two time out.

Shown: structured error context. Choose what the subagent returns, and follow it to the report.

Suppressed error. The coordinator was told all six searches succeeded, so it never retries, and the report states two falsehoods with confidence.

Generic status. The local retries were right; throwing away the failure type, the queries, the four markets and the alternatives was not.

Terminated on a single failure. Two unreachable markets cost everything that worked.

Structured error context. The coordinator reroutes the two markets, recovers Italy, and the report discloses the one market still missing.

Time reads downward. The retries happen inside the subagent's own context window, so the middle box is all the coordinator knows. Only the structured report lets it reroute the two markets, recover Italy and disclose Poland.

KnowledgeWhich agent retries

Recovery splits by what the retry needs. The subagent retries when the identical call can succeed after a short wait, as with a transient timeout or rate limit, because the coordinator gains nothing from hearing about a blip and every report costs it context. The retries are bounded, here two per query, so a real outage still surfaces quickly. The coordinator owns the next attempt when recovery needs something the subagent lacks: another source or subagent, a narrower scope, or a decision to ship with a gap. A coordinator that re-sends the query the subagent already retried is repeating work, which is why the report must say what was attempted.

KnowledgeTwo replies that look alike and mean opposite things

Italy's query never completed, so nothing is known about Italian incentives, and the coordinator must decide whether to retry, reroute or disclose. Portugal's query ran and matched nothing, which is a finished, correct answer, so "no national programme found" goes in the report as a finding. If the subagent returns both as an empty list marked success, the coordinator treats Italy as a finding too, and that silent suppression makes every later agent repeat the false absence with confidence.

Access failure

The source was never searched

{ "market": "IT",
  "status": "error",
  "failure_type": "timeout",
  "attempted": "incentives IT, 2 retries" }

Needs a decision: retry differently, reroute, or proceed and disclose.

Valid empty result

The source was searched and had nothing

{ "market": "PT",
  "status": "success",
  "results": [] }

Needs nothing: "no programme found" is a finding.

SkillFrom a false absence to a disclosed gap

The first draft says Italy and Poland offer no national storage incentives, and a reviewer knows both do. The web search subagent had caught the two timeouts and returned empty lists as success, so every later agent treated an access failure as a result. The fix keeps the four markets that worked and adds structured error context beside them.

// web search subagent → coordinator, after local retries
{ "status": "partial",
  "results": { "DE": [ … ], "FR": [ … ], "ES": [ … ], "PT": [] },
  "errors": [ { "failure_type": "timeout", "markets": ["IT", "PL"],
      "attempted": "incentives query per market, 2 retries each",
      "alternatives": ["national energy agency registries"] } ] }

The coordinator re-delegates Italy and Poland to the document analysis subagent with the registries as sources. Italy's registry answers and Poland's is offline, so the coordinator proceeds instead of terminating the workflow. The synthesis subagent adds coverage annotations: which findings are well supported and which topic areas have gaps because sources were unavailable.

## Coverage (synthesis subagent)
Well supported  cost trends (9 sources) · deployments (6 sources)
Supported       incentives, 5 of 6 markets (PT: no programme exists)
Gap             incentives, Poland: source unreachable after retries
                and a registry fallback; absent, not zero
For each failure, ask: Who can recover from it, and with what?
The same call should work after a short waitThe subagent retries, a bounded number of times, and says nothing if it clears.
It persists, or needs another source or scopeThe subagent returns failure type, attempt, partial results and alternatives. The coordinator reroutes or narrows.
No route reaches the sourceThe coordinator proceeds, and the synthesis output annotates the gap.
Takeaway
  • A subagent's final message is the coordinator's only view of a failure, so a generic status leaves it guessing.
  • "Timed out" and "searched, found nothing" are different results. Sending both as an empty success turns an outage into a false finding.
  • One unreachable source should cost a disclosed gap in the report, never the whole workflow.

check yourself

0 / 3 answered

The document analysis subagent must summarise twelve journal papers on sodium-ion cells. The publisher's API returns HTTP 403 for six of them on every call, because the team's licence does not cover that journal. Open preprints of five of those six are listed on a public preprint server, which only the web search subagent can reach.Which design most directly lets the system recover the missing papers?

Select one.

A report on EV charging adoption covers five regions. One region's statistics portal failed DNS lookups all week, the document analysis subagent returned {"stats": []} for it as a success, and the report reads as complete. The team still wants each report delivered on schedule when one source is down.Which TWO changes best prevent a silent gap while keeping delivery on schedule?

Select two.

The coordinator's log shows the web search subagent returned {"status": "error", "message": "search unavailable"} for the solar-tariff topic. The coordinator re-sent the same topic three times, got the same reply each time, and dropped it. The subagent's own trace shows 7 of 9 queries succeeded; the other 2 hit HTTP 404 on a retired government URL.What is the most likely root cause of the lost topic?

Select one.

sources

  1. Anthropic Engineering, "How we built our multi-agent research system" — a lead agent delegating to subagents; "letting the agent know when a tool is failing and letting it adapt works surprisingly well"; retry logic and checkpoints rather than restarting from the beginning. Checked 2026-10-01.
  2. Claude Docs, "Subagents in the SDK" — each subagent runs in its own conversation, and "only its final message returns to the parent". Checked 2026-10-01.
  3. The four parts of structured error context, access failures versus valid empty results, "search unavailable", local recovery, coverage annotations, and the two anti-patterns: Claude Certified Architect — Foundations Exam Guide v1.0, Domain 5.3. Scenario: the guide's Multi-Agent Research System. The battery-storage report and its six markets are this course's illustration.